2 * Copyright (c) 1982, 1986, 1988, 1990, 1993
3 * The Regents of the University of California. All rights reserved.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 3. Neither the name of the University nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94
30 * $FreeBSD: src/sys/netinet/ip_output.c,v 1.99.2.37 2003/04/15 06:44:45 silby Exp $
36 #include "opt_ipdivert.h"
37 #include "opt_ipsec.h"
38 #include "opt_mbuf_stress_test.h"
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/kernel.h>
44 #include <sys/malloc.h>
46 #include <sys/protosw.h>
47 #include <sys/socket.h>
48 #include <sys/socketvar.h>
51 #include <sys/sysctl.h>
52 #include <sys/in_cksum.h>
55 #include <sys/thread2.h>
56 #include <sys/mplock2.h>
57 #include <sys/msgport2.h>
60 #include <net/netisr.h>
62 #include <net/route.h>
64 #include <netinet/in.h>
65 #include <netinet/in_systm.h>
66 #include <netinet/ip.h>
67 #include <netinet/in_pcb.h>
68 #include <netinet/in_var.h>
69 #include <netinet/ip_var.h>
71 #include <netproto/mpls/mpls_var.h>
73 static MALLOC_DEFINE(M_IPMOPTS
, "ip_moptions", "internet multicast options");
76 #include <netinet6/ipsec.h>
77 #include <netproto/key/key.h>
79 #include <netproto/key/key_debug.h>
81 #define KEYDEBUG(lev,arg)
86 #include <netproto/ipsec/ipsec.h>
87 #include <netproto/ipsec/xform.h>
88 #include <netproto/ipsec/key.h>
91 #include <net/ipfw/ip_fw.h>
92 #include <net/dummynet/ip_dummynet.h>
94 #define print_ip(x, a, y) kprintf("%s %d.%d.%d.%d%s",\
95 x, (ntohl(a.s_addr)>>24)&0xFF,\
96 (ntohl(a.s_addr)>>16)&0xFF,\
97 (ntohl(a.s_addr)>>8)&0xFF,\
98 (ntohl(a.s_addr))&0xFF, y);
102 #ifdef MBUF_STRESS_TEST
103 int mbuf_frag_size
= 0;
104 SYSCTL_INT(_net_inet_ip
, OID_AUTO
, mbuf_frag_size
, CTLFLAG_RW
,
105 &mbuf_frag_size
, 0, "Fragment outgoing mbufs to this size");
108 static struct mbuf
*ip_insertoptions(struct mbuf
*, struct mbuf
*, int *);
109 static struct ifnet
*ip_multicast_if(struct in_addr
*, int *);
110 static void ip_mloopback
111 (struct ifnet
*, struct mbuf
*, struct sockaddr_in
*, int);
112 static int ip_getmoptions
113 (struct sockopt
*, struct ip_moptions
*);
114 static int ip_pcbopts(int, struct mbuf
**, struct mbuf
*);
115 static int ip_setmoptions
116 (struct sockopt
*, struct ip_moptions
**);
118 int ip_optcopy(struct ip
*, struct ip
*);
120 extern int route_assert_owner_access
;
122 extern struct protosw inetsw
[];
125 ip_localforward(struct mbuf
*m
, const struct sockaddr_in
*dst
, int hlen
)
127 struct in_ifaddr_container
*iac
;
130 * We need to figure out if we have been forwarded to a local
131 * socket. If so, then we should somehow "loop back" to
132 * ip_input(), and get directed to the PCB as if we had received
133 * this packet. This is because it may be difficult to identify
134 * the packets you want to forward until they are being output
135 * and have selected an interface (e.g. locally initiated
136 * packets). If we used the loopback inteface, we would not be
137 * able to control what happens as the packet runs through
138 * ip_input() as it is done through a ISR.
140 LIST_FOREACH(iac
, INADDR_HASH(dst
->sin_addr
.s_addr
), ia_hash
) {
142 * If the addr to forward to is one of ours, we pretend
143 * to be the destination for this packet.
145 if (IA_SIN(iac
->ia
)->sin_addr
.s_addr
== dst
->sin_addr
.s_addr
)
151 if (m
->m_pkthdr
.rcvif
== NULL
)
152 m
->m_pkthdr
.rcvif
= ifunit("lo0");
153 if (m
->m_pkthdr
.csum_flags
& CSUM_DELAY_DATA
) {
154 m
->m_pkthdr
.csum_flags
|= CSUM_DATA_VALID
|
156 m
->m_pkthdr
.csum_data
= 0xffff;
158 m
->m_pkthdr
.csum_flags
|= CSUM_IP_CHECKED
| CSUM_IP_VALID
;
161 * Make sure that the IP header is in one mbuf,
162 * required by ip_input
164 if (m
->m_len
< hlen
) {
165 m
= m_pullup(m
, hlen
);
167 /* The packet was freed; we are done */
171 ip
= mtod(m
, struct ip
*);
173 ip
->ip_len
= htons(ip
->ip_len
);
174 ip
->ip_off
= htons(ip
->ip_off
);
177 return 1; /* The packet gets forwarded locally */
183 * IP output. The packet in mbuf chain m contains a skeletal IP
184 * header (with len, off, ttl, proto, tos, src, dst).
185 * The mbuf chain containing the packet will be freed.
186 * The mbuf opt, if present, will not be freed.
189 ip_output(struct mbuf
*m0
, struct mbuf
*opt
, struct route
*ro
,
190 int flags
, struct ip_moptions
*imo
, struct inpcb
*inp
)
193 struct ifnet
*ifp
= NULL
; /* keep compiler happy */
195 int hlen
= sizeof(struct ip
);
197 struct sockaddr_in
*dst
= NULL
; /* keep compiler happy */
198 struct in_ifaddr
*ia
= NULL
;
199 int isbroadcast
, sw_csum
;
200 struct in_addr pkt_dst
;
201 struct route iproute
;
204 struct secpolicy
*sp
= NULL
;
205 struct socket
*so
= inp
? inp
->inp_socket
: NULL
;
208 struct secpolicy
*sp
= NULL
;
209 struct tdb_ident
*tdbi
;
210 #endif /* FAST_IPSEC */
211 struct sockaddr_in
*next_hop
= NULL
;
212 int src_was_INADDR_ANY
= 0; /* as the name says... */
219 bzero(ro
, sizeof *ro
);
220 } else if (ro
->ro_rt
!= NULL
&& ro
->ro_rt
->rt_cpuid
!= mycpuid
) {
221 if (flags
& IP_DEBUGROUTE
) {
222 if (route_assert_owner_access
) {
224 "rt rt_cpuid %d accessed on cpu %d\n",
225 ro
->ro_rt
->rt_cpuid
, mycpuid
);
227 kprintf("ip_output: "
228 "rt rt_cpuid %d accessed on cpu %d\n",
229 ro
->ro_rt
->rt_cpuid
, mycpuid
);
236 * If the cached rtentry's owner CPU is not the current CPU,
237 * then don't touch the cached rtentry (remote free is too
238 * expensive in this context); just relocate the route.
241 bzero(ro
, sizeof *ro
);
244 if (m
->m_pkthdr
.fw_flags
& IPFORWARD_MBUF_TAGGED
) {
246 mtag
= m_tag_find(m
, PACKET_TAG_IPFORWARD
, NULL
);
247 KKASSERT(mtag
!= NULL
);
248 next_hop
= m_tag_data(mtag
);
251 if (m
->m_pkthdr
.fw_flags
& DUMMYNET_MBUF_TAGGED
) {
252 struct dn_pkt
*dn_pkt
;
254 /* Extract info from dummynet tag */
255 mtag
= m_tag_find(m
, PACKET_TAG_DUMMYNET
, NULL
);
256 KKASSERT(mtag
!= NULL
);
257 dn_pkt
= m_tag_data(mtag
);
260 * The packet was already tagged, so part of the
261 * processing was already done, and we need to go down.
262 * Get the calculated parameters from the tag.
266 KKASSERT(ro
== &iproute
);
267 *ro
= dn_pkt
->ro
; /* structure copy */
268 KKASSERT(ro
->ro_rt
== NULL
|| ro
->ro_rt
->rt_cpuid
== mycpuid
);
270 dst
= dn_pkt
->dn_dst
;
271 if (dst
== (struct sockaddr_in
*)&(dn_pkt
->ro
.ro_dst
)) {
272 /* If 'dst' points into dummynet tag, adjust it */
273 dst
= (struct sockaddr_in
*)&(ro
->ro_dst
);
276 ip
= mtod(m
, struct ip
*);
277 hlen
= IP_VHL_HL(ip
->ip_vhl
) << 2 ;
279 ia
= ifatoia(ro
->ro_rt
->rt_ifa
);
285 m
= ip_insertoptions(m
, opt
, &len
);
289 ip
= mtod(m
, struct ip
*);
294 if (!(flags
& (IP_FORWARDING
|IP_RAWOUTPUT
))) {
295 ip
->ip_vhl
= IP_MAKE_VHL(IPVERSION
, hlen
>> 2);
297 ip
->ip_id
= ip_newid();
298 ipstat
.ips_localout
++;
300 hlen
= IP_VHL_HL(ip
->ip_vhl
) << 2;
304 pkt_dst
= next_hop
? next_hop
->sin_addr
: ip
->ip_dst
;
306 dst
= (struct sockaddr_in
*)&ro
->ro_dst
;
308 * If there is a cached route,
309 * check that it is to the same destination
310 * and is still up. If not, free it and try again.
311 * The address family should also be checked in case of sharing the
315 (!(ro
->ro_rt
->rt_flags
& RTF_UP
) ||
316 dst
->sin_family
!= AF_INET
||
317 dst
->sin_addr
.s_addr
!= pkt_dst
.s_addr
)) {
321 if (ro
->ro_rt
== NULL
) {
322 bzero(dst
, sizeof *dst
);
323 dst
->sin_family
= AF_INET
;
324 dst
->sin_len
= sizeof *dst
;
325 dst
->sin_addr
= pkt_dst
;
328 * If routing to interface only,
329 * short circuit routing lookup.
331 if (flags
& IP_ROUTETOIF
) {
332 if ((ia
= ifatoia(ifa_ifwithdstaddr(sintosa(dst
)))) == NULL
&&
333 (ia
= ifatoia(ifa_ifwithnet(sintosa(dst
)))) == NULL
) {
334 ipstat
.ips_noroute
++;
340 isbroadcast
= in_broadcast(dst
->sin_addr
, ifp
);
341 } else if (IN_MULTICAST(ntohl(pkt_dst
.s_addr
)) &&
342 imo
!= NULL
&& imo
->imo_multicast_ifp
!= NULL
) {
344 * Bypass the normal routing lookup for multicast
345 * packets if the interface is specified.
347 ifp
= imo
->imo_multicast_ifp
;
349 isbroadcast
= 0; /* fool gcc */
352 * If this is the case, we probably don't want to allocate
353 * a protocol-cloned route since we didn't get one from the
354 * ULP. This lets TCP do its thing, while not burdening
355 * forwarding or ICMP with the overhead of cloning a route.
356 * Of course, we still want to do any cloning requested by
357 * the link layer, as this is probably required in all cases
358 * for correct operation (as it is for ARP).
360 if (ro
->ro_rt
== NULL
)
361 rtalloc_ign(ro
, RTF_PRCLONING
);
362 if (ro
->ro_rt
== NULL
) {
363 ipstat
.ips_noroute
++;
364 error
= EHOSTUNREACH
;
367 ia
= ifatoia(ro
->ro_rt
->rt_ifa
);
368 ifp
= ro
->ro_rt
->rt_ifp
;
370 if (ro
->ro_rt
->rt_flags
& RTF_GATEWAY
)
371 dst
= (struct sockaddr_in
*)ro
->ro_rt
->rt_gateway
;
372 if (ro
->ro_rt
->rt_flags
& RTF_HOST
)
373 isbroadcast
= (ro
->ro_rt
->rt_flags
& RTF_BROADCAST
);
375 isbroadcast
= in_broadcast(dst
->sin_addr
, ifp
);
377 if (IN_MULTICAST(ntohl(pkt_dst
.s_addr
))) {
378 struct in_multi
*inm
;
380 m
->m_flags
|= M_MCAST
;
382 * IP destination address is multicast. Make sure "dst"
383 * still points to the address in "ro". (It may have been
384 * changed to point to a gateway address, above.)
386 dst
= (struct sockaddr_in
*)&ro
->ro_dst
;
388 * See if the caller provided any multicast options
391 ip
->ip_ttl
= imo
->imo_multicast_ttl
;
392 if (imo
->imo_multicast_vif
!= -1) {
395 ip_mcast_src(imo
->imo_multicast_vif
) :
399 ip
->ip_ttl
= IP_DEFAULT_MULTICAST_TTL
;
402 * Confirm that the outgoing interface supports multicast.
404 if ((imo
== NULL
) || (imo
->imo_multicast_vif
== -1)) {
405 if (!(ifp
->if_flags
& IFF_MULTICAST
)) {
406 ipstat
.ips_noroute
++;
412 * If source address not specified yet, use address
413 * of outgoing interface.
415 if (ip
->ip_src
.s_addr
== INADDR_ANY
) {
416 /* Interface may have no addresses. */
418 ip
->ip_src
= IA_SIN(ia
)->sin_addr
;
421 IN_LOOKUP_MULTI(pkt_dst
, ifp
, inm
);
423 (imo
== NULL
|| imo
->imo_multicast_loop
)) {
425 * If we belong to the destination multicast group
426 * on the outgoing interface, and the caller did not
427 * forbid loopback, loop back a copy.
429 ip_mloopback(ifp
, m
, dst
, hlen
);
432 * If we are acting as a multicast router, perform
433 * multicast forwarding as if the packet had just
434 * arrived on the interface to which we are about
435 * to send. The multicast forwarding function
436 * recursively calls this function, using the
437 * IP_FORWARDING flag to prevent infinite recursion.
439 * Multicasts that are looped back by ip_mloopback(),
440 * above, will be forwarded by the ip_input() routine,
443 if (ip_mrouter
&& !(flags
& IP_FORWARDING
)) {
445 * If rsvp daemon is not running, do not
446 * set ip_moptions. This ensures that the packet
447 * is multicast and not just sent down one link
448 * as prescribed by rsvpd.
454 if (ip_mforward(ip
, ifp
, m
, imo
) != 0) {
465 * Multicasts with a time-to-live of zero may be looped-
466 * back, above, but must not be transmitted on a network.
467 * Also, multicasts addressed to the loopback interface
468 * are not sent -- the above call to ip_mloopback() will
469 * loop back a copy if this host actually belongs to the
470 * destination group on the loopback interface.
472 if (ip
->ip_ttl
== 0 || ifp
->if_flags
& IFF_LOOPBACK
) {
479 m
->m_flags
&= ~M_MCAST
;
483 * If the source address is not specified yet, use the address
484 * of the outoing interface. In case, keep note we did that, so
485 * if the the firewall changes the next-hop causing the output
486 * interface to change, we can fix that.
488 if (ip
->ip_src
.s_addr
== INADDR_ANY
|| src_was_INADDR_ANY
) {
489 /* Interface may have no addresses. */
491 ip
->ip_src
= IA_SIN(ia
)->sin_addr
;
492 src_was_INADDR_ANY
= 1;
498 * Disable packet drop hack.
499 * Packetdrop should be done by queueing.
503 * Verify that we have any chance at all of being able to queue
504 * the packet or packet fragments
506 if ((ifp
->if_snd
.ifq_len
+ ip
->ip_len
/ ifp
->if_mtu
+ 1) >=
507 ifp
->if_snd
.ifq_maxlen
) {
509 ipstat
.ips_odropped
++;
515 * Look for broadcast address and
516 * verify user is allowed to send
520 if (!(ifp
->if_flags
& IFF_BROADCAST
)) {
521 error
= EADDRNOTAVAIL
;
524 if (!(flags
& IP_ALLOWBROADCAST
)) {
528 /* don't allow broadcast messages to be fragmented */
529 if (ip
->ip_len
> ifp
->if_mtu
) {
533 m
->m_flags
|= M_BCAST
;
535 m
->m_flags
&= ~M_BCAST
;
540 /* get SP for this packet */
542 sp
= ipsec4_getpolicybyaddr(m
, IPSEC_DIR_OUTBOUND
, flags
, &error
);
544 sp
= ipsec4_getpolicybysock(m
, IPSEC_DIR_OUTBOUND
, so
, &error
);
547 ipsecstat
.out_inval
++;
554 switch (sp
->policy
) {
555 case IPSEC_POLICY_DISCARD
:
557 * This packet is just discarded.
559 ipsecstat
.out_polvio
++;
562 case IPSEC_POLICY_BYPASS
:
563 case IPSEC_POLICY_NONE
:
564 case IPSEC_POLICY_TCP
:
565 /* no need to do IPsec. */
568 case IPSEC_POLICY_IPSEC
:
569 if (sp
->req
== NULL
) {
570 /* acquire a policy */
571 error
= key_spdacquire(sp
);
576 case IPSEC_POLICY_ENTRUST
:
578 kprintf("ip_output: Invalid policy found. %d\n", sp
->policy
);
581 struct ipsec_output_state state
;
582 bzero(&state
, sizeof state
);
584 if (flags
& IP_ROUTETOIF
) {
586 bzero(&iproute
, sizeof iproute
);
589 state
.dst
= (struct sockaddr
*)dst
;
595 * delayed checksums are not currently compatible with IPsec
597 if (m
->m_pkthdr
.csum_flags
& CSUM_DELAY_DATA
) {
599 m
->m_pkthdr
.csum_flags
&= ~CSUM_DELAY_DATA
;
602 ip
->ip_len
= htons(ip
->ip_len
);
603 ip
->ip_off
= htons(ip
->ip_off
);
605 error
= ipsec4_output(&state
, sp
, flags
);
608 if (flags
& IP_ROUTETOIF
) {
610 * if we have tunnel mode SA, we may need to ignore
613 if (state
.ro
!= &iproute
|| state
.ro
->ro_rt
!= NULL
) {
614 flags
&= ~IP_ROUTETOIF
;
619 dst
= (struct sockaddr_in
*)state
.dst
;
621 /* mbuf is already reclaimed in ipsec4_output. */
631 kprintf("ip4_output (ipsec): error code %d\n", error
);
634 /* don't show these error codes to the user */
642 /* be sure to update variables that are affected by ipsec4_output() */
643 ip
= mtod(m
, struct ip
*);
645 hlen
= IP_VHL_HL(ip
->ip_vhl
) << 2;
647 hlen
= ip
->ip_hl
<< 2;
649 if (ro
->ro_rt
== NULL
) {
650 if (!(flags
& IP_ROUTETOIF
)) {
651 kprintf("ip_output: "
652 "can't update route after IPsec processing\n");
653 error
= EHOSTUNREACH
; /*XXX*/
657 ia
= ifatoia(ro
->ro_rt
->rt_ifa
);
658 ifp
= ro
->ro_rt
->rt_ifp
;
661 /* make it flipped, again. */
662 ip
->ip_len
= ntohs(ip
->ip_len
);
663 ip
->ip_off
= ntohs(ip
->ip_off
);
668 * Check the security policy (SP) for the packet and, if
669 * required, do IPsec-related processing. There are two
670 * cases here; the first time a packet is sent through
671 * it will be untagged and handled by ipsec4_checkpolicy.
672 * If the packet is resubmitted to ip_output (e.g. after
673 * AH, ESP, etc. processing), there will be a tag to bypass
674 * the lookup and related policy checking.
676 mtag
= m_tag_find(m
, PACKET_TAG_IPSEC_PENDING_TDB
, NULL
);
679 tdbi
= (struct tdb_ident
*)m_tag_data(mtag
);
680 sp
= ipsec_getpolicy(tdbi
, IPSEC_DIR_OUTBOUND
);
682 error
= -EINVAL
; /* force silent drop */
683 m_tag_delete(m
, mtag
);
685 sp
= ipsec4_checkpolicy(m
, IPSEC_DIR_OUTBOUND
, flags
,
689 * There are four return cases:
690 * sp != NULL apply IPsec policy
691 * sp == NULL, error == 0 no IPsec handling needed
692 * sp == NULL, error == -EINVAL discard packet w/o error
693 * sp == NULL, error != 0 discard packet, report error
696 /* Loop detection, check if ipsec processing already done */
697 KASSERT(sp
->req
!= NULL
, ("ip_output: no ipsec request"));
698 for (mtag
= m_tag_first(m
); mtag
!= NULL
;
699 mtag
= m_tag_next(m
, mtag
)) {
700 if (mtag
->m_tag_cookie
!= MTAG_ABI_COMPAT
)
702 if (mtag
->m_tag_id
!= PACKET_TAG_IPSEC_OUT_DONE
&&
703 mtag
->m_tag_id
!= PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED
)
706 * Check if policy has an SA associated with it.
707 * This can happen when an SP has yet to acquire
708 * an SA; e.g. on first reference. If it occurs,
709 * then we let ipsec4_process_packet do its thing.
711 if (sp
->req
->sav
== NULL
)
713 tdbi
= (struct tdb_ident
*)m_tag_data(mtag
);
714 if (tdbi
->spi
== sp
->req
->sav
->spi
&&
715 tdbi
->proto
== sp
->req
->sav
->sah
->saidx
.proto
&&
716 bcmp(&tdbi
->dst
, &sp
->req
->sav
->sah
->saidx
.dst
,
717 sizeof(union sockaddr_union
)) == 0) {
719 * No IPsec processing is needed, free
722 * NB: null pointer to avoid free at
725 KEY_FREESP(&sp
), sp
= NULL
;
732 * Do delayed checksums now because we send before
733 * this is done in the normal processing path.
735 if (m
->m_pkthdr
.csum_flags
& CSUM_DELAY_DATA
) {
737 m
->m_pkthdr
.csum_flags
&= ~CSUM_DELAY_DATA
;
740 ip
->ip_len
= htons(ip
->ip_len
);
741 ip
->ip_off
= htons(ip
->ip_off
);
743 /* NB: callee frees mbuf */
744 error
= ipsec4_process_packet(m
, sp
->req
, flags
, 0);
746 * Preserve KAME behaviour: ENOENT can be returned
747 * when an SA acquire is in progress. Don't propagate
748 * this to user-level; it confuses applications.
750 * XXX this will go away when the SADB is redone.
761 * Hack: -EINVAL is used to signal that a packet
762 * should be silently discarded. This is typically
763 * because we asked key management for an SA and
764 * it was delayed (e.g. kicked up to IKE).
766 if (error
== -EINVAL
)
770 /* No IPsec processing for this packet. */
774 * If deferred crypto processing is needed, check that
775 * the interface supports it.
777 mtag
= m_tag_find(m
, PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED
, NULL
);
778 if (mtag
!= NULL
&& !(ifp
->if_capenable
& IFCAP_IPSEC
)) {
779 /* notify IPsec to do its own crypto */
780 ipsp_skipcrypto_unmark((struct tdb_ident
*)m_tag_data(mtag
));
781 error
= EHOSTUNREACH
;
787 #endif /* FAST_IPSEC */
789 /* We are already being fwd'd from a firewall. */
790 if (next_hop
!= NULL
)
794 if (!pfil_has_hooks(&inet_pfil_hook
)) {
795 if (m
->m_pkthdr
.fw_flags
& DUMMYNET_MBUF_TAGGED
) {
797 * Strip dummynet tags from stranded packets
799 mtag
= m_tag_find(m
, PACKET_TAG_DUMMYNET
, NULL
);
800 KKASSERT(mtag
!= NULL
);
801 m_tag_delete(m
, mtag
);
802 m
->m_pkthdr
.fw_flags
&= ~DUMMYNET_MBUF_TAGGED
;
809 * - Xlate: translate packet's addr/port (NAT).
810 * - Firewall: deny/allow/etc.
811 * - Wrap: fake packet's addr/port <unimpl.>
812 * - Encapsulate: put it in another IP and send out. <unimp.>
816 * Run through list of hooks for output packets.
818 error
= pfil_run_hooks(&inet_pfil_hook
, &m
, ifp
, PFIL_OUT
);
819 if (error
!= 0 || m
== NULL
)
821 ip
= mtod(m
, struct ip
*);
823 if (m
->m_pkthdr
.fw_flags
& IPFORWARD_MBUF_TAGGED
) {
825 * Check dst to make sure it is directly reachable on the
826 * interface we previously thought it was.
827 * If it isn't (which may be likely in some situations) we have
828 * to re-route it (ie, find a route for the next-hop and the
829 * associated interface) and set them here. This is nested
830 * forwarding which in most cases is undesirable, except where
831 * such control is nigh impossible. So we do it here.
834 mtag
= m_tag_find(m
, PACKET_TAG_IPFORWARD
, NULL
);
835 KKASSERT(mtag
!= NULL
);
836 next_hop
= m_tag_data(mtag
);
839 * Try local forwarding first
841 if (ip_localforward(m
, next_hop
, hlen
))
845 * Relocate the route based on next_hop.
846 * If the current route is inp's cache, keep it untouched.
848 if (ro
== &iproute
&& ro
->ro_rt
!= NULL
) {
853 bzero(ro
, sizeof *ro
);
856 * Forwarding to broadcast address is not allowed.
857 * XXX Should we follow IP_ROUTETOIF?
859 flags
&= ~(IP_ALLOWBROADCAST
| IP_ROUTETOIF
);
861 /* We are doing forwarding now */
862 flags
|= IP_FORWARDING
;
867 if (m
->m_pkthdr
.fw_flags
& DUMMYNET_MBUF_TAGGED
) {
868 struct dn_pkt
*dn_pkt
;
870 mtag
= m_tag_find(m
, PACKET_TAG_DUMMYNET
, NULL
);
871 KKASSERT(mtag
!= NULL
);
872 dn_pkt
= m_tag_data(mtag
);
875 * Under certain cases it is not possible to recalculate
876 * 'ro' and 'dst', let alone 'flags', so just save them in
877 * dummynet tag and avoid the possible wrong reculcalation
878 * when we come back to ip_output() again.
880 * All other parameters have been already used and so they
881 * are not needed anymore.
882 * XXX if the ifp is deleted while a pkt is in dummynet,
883 * we are in trouble! (TODO use ifnet_detach_event)
885 * We need to copy *ro because for ICMP pkts (and maybe
886 * others) the caller passed a pointer into the stack;
887 * dst might also be a pointer into *ro so it needs to
892 ro
->ro_rt
->rt_refcnt
++;
893 if (dst
== (struct sockaddr_in
*)&ro
->ro_dst
) {
894 /* 'dst' points into 'ro' */
895 dst
= (struct sockaddr_in
*)&(dn_pkt
->ro
.ro_dst
);
897 dn_pkt
->dn_dst
= dst
;
898 dn_pkt
->flags
= flags
;
904 /* 127/8 must not appear on wire - RFC1122. */
905 if ((ntohl(ip
->ip_dst
.s_addr
) >> IN_CLASSA_NSHIFT
) == IN_LOOPBACKNET
||
906 (ntohl(ip
->ip_src
.s_addr
) >> IN_CLASSA_NSHIFT
) == IN_LOOPBACKNET
) {
907 if (!(ifp
->if_flags
& IFF_LOOPBACK
)) {
908 ipstat
.ips_badaddr
++;
909 error
= EADDRNOTAVAIL
;
914 m
->m_pkthdr
.csum_flags
|= CSUM_IP
;
915 sw_csum
= m
->m_pkthdr
.csum_flags
& ~ifp
->if_hwassist
;
916 if (sw_csum
& CSUM_DELAY_DATA
) {
918 sw_csum
&= ~CSUM_DELAY_DATA
;
920 m
->m_pkthdr
.csum_flags
&= ifp
->if_hwassist
;
923 * If small enough for interface, or the interface will take
924 * care of the fragmentation for us, can just send directly.
926 if (ip
->ip_len
<= ifp
->if_mtu
|| ((ifp
->if_hwassist
& CSUM_FRAGMENT
) &&
927 !(ip
->ip_off
& IP_DF
))) {
928 ip
->ip_len
= htons(ip
->ip_len
);
929 ip
->ip_off
= htons(ip
->ip_off
);
931 if (sw_csum
& CSUM_DELAY_IP
) {
932 if (ip
->ip_vhl
== IP_VHL_BORING
)
933 ip
->ip_sum
= in_cksum_hdr(ip
);
935 ip
->ip_sum
= in_cksum(m
, hlen
);
938 /* Record statistics for this interface address. */
939 if (!(flags
& IP_FORWARDING
) && ia
) {
940 ia
->ia_ifa
.if_opackets
++;
941 ia
->ia_ifa
.if_obytes
+= m
->m_pkthdr
.len
;
945 /* clean ipsec history once it goes out of the node */
949 #ifdef MBUF_STRESS_TEST
950 if (mbuf_frag_size
&& m
->m_pkthdr
.len
> mbuf_frag_size
) {
951 struct mbuf
*m1
, *m2
;
954 tmp
= length
= m
->m_pkthdr
.len
;
956 while ((length
-= mbuf_frag_size
) >= 1) {
957 m1
= m_split(m
, length
, MB_DONTWAIT
);
961 while (m2
->m_next
!= NULL
)
965 m
->m_pkthdr
.len
= tmp
;
970 if (!mpls_output_process(m
, ro
->ro_rt
))
973 error
= ifp
->if_output(ifp
, m
, (struct sockaddr
*)dst
,
978 if (ip
->ip_off
& IP_DF
) {
981 * This case can happen if the user changed the MTU
982 * of an interface after enabling IP on it. Because
983 * most netifs don't keep track of routes pointing to
984 * them, there is no way for one to update all its
985 * routes when the MTU is changed.
987 if ((ro
->ro_rt
->rt_flags
& (RTF_UP
| RTF_HOST
)) &&
988 !(ro
->ro_rt
->rt_rmx
.rmx_locks
& RTV_MTU
) &&
989 (ro
->ro_rt
->rt_rmx
.rmx_mtu
> ifp
->if_mtu
)) {
990 ro
->ro_rt
->rt_rmx
.rmx_mtu
= ifp
->if_mtu
;
992 ipstat
.ips_cantfrag
++;
997 * Too large for interface; fragment if possible. If successful,
998 * on return, m will point to a list of packets to be sent.
1000 error
= ip_fragment(ip
, &m
, ifp
->if_mtu
, ifp
->if_hwassist
, sw_csum
);
1005 m
->m_nextpkt
= NULL
;
1007 /* clean ipsec history once it goes out of the node */
1011 /* Record statistics for this interface address. */
1013 ia
->ia_ifa
.if_opackets
++;
1014 ia
->ia_ifa
.if_obytes
+= m
->m_pkthdr
.len
;
1017 if (!mpls_output_process(m
, ro
->ro_rt
))
1020 error
= ifp
->if_output(ifp
, m
, (struct sockaddr
*)dst
,
1028 ipstat
.ips_fragmented
++;
1031 if (ro
== &iproute
&& ro
->ro_rt
!= NULL
) {
1037 KEYDEBUG(KEYDEBUG_IPSEC_STAMP
,
1038 kprintf("DP ip_output call free SP:%p\n", sp
));
1053 * Create a chain of fragments which fit the given mtu. m_frag points to the
1054 * mbuf to be fragmented; on return it points to the chain with the fragments.
1055 * Return 0 if no error. If error, m_frag may contain a partially built
1056 * chain of fragments that should be freed by the caller.
1058 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
1059 * sw_csum contains the delayed checksums flags (e.g., CSUM_DELAY_IP).
1062 ip_fragment(struct ip
*ip
, struct mbuf
**m_frag
, int mtu
,
1063 u_long if_hwassist_flags
, int sw_csum
)
1066 int hlen
= IP_VHL_HL(ip
->ip_vhl
) << 2;
1067 int len
= (mtu
- hlen
) & ~7; /* size of payload in each fragment */
1069 struct mbuf
*m0
= *m_frag
; /* the original packet */
1071 struct mbuf
**mnext
;
1074 if (ip
->ip_off
& IP_DF
) { /* Fragmentation not allowed */
1075 ipstat
.ips_cantfrag
++;
1080 * Must be able to put at least 8 bytes per fragment.
1086 * If the interface will not calculate checksums on
1087 * fragmented packets, then do it here.
1089 if ((m0
->m_pkthdr
.csum_flags
& CSUM_DELAY_DATA
) &&
1090 !(if_hwassist_flags
& CSUM_IP_FRAGS
)) {
1091 in_delayed_cksum(m0
);
1092 m0
->m_pkthdr
.csum_flags
&= ~CSUM_DELAY_DATA
;
1095 if (len
> PAGE_SIZE
) {
1097 * Fragment large datagrams such that each segment
1098 * contains a multiple of PAGE_SIZE amount of data,
1099 * plus headers. This enables a receiver to perform
1100 * page-flipping zero-copy optimizations.
1102 * XXX When does this help given that sender and receiver
1103 * could have different page sizes, and also mtu could
1104 * be less than the receiver's page size ?
1109 for (m
= m0
, off
= 0; m
&& (off
+m
->m_len
) <= mtu
; m
= m
->m_next
)
1113 * firstlen (off - hlen) must be aligned on an
1117 goto smart_frag_failure
;
1118 off
= ((off
- hlen
) & ~7) + hlen
;
1119 newlen
= (~PAGE_MASK
) & mtu
;
1120 if ((newlen
+ sizeof(struct ip
)) > mtu
) {
1121 /* we failed, go back the default */
1132 firstlen
= off
- hlen
;
1133 mnext
= &m0
->m_nextpkt
; /* pointer to next packet */
1136 * Loop through length of segment after first fragment,
1137 * make new header and copy data of each part and link onto chain.
1138 * Here, m0 is the original packet, m is the fragment being created.
1139 * The fragments are linked off the m_nextpkt of the original
1140 * packet, which after processing serves as the first fragment.
1142 for (nfrags
= 1; off
< ip
->ip_len
; off
+= len
, nfrags
++) {
1143 struct ip
*mhip
; /* ip header on the fragment */
1145 int mhlen
= sizeof(struct ip
);
1147 MGETHDR(m
, MB_DONTWAIT
, MT_HEADER
);
1150 ipstat
.ips_odropped
++;
1153 m
->m_flags
|= (m0
->m_flags
& M_MCAST
) | M_FRAG
;
1155 * In the first mbuf, leave room for the link header, then
1156 * copy the original IP header including options. The payload
1157 * goes into an additional mbuf chain returned by m_copy().
1159 m
->m_data
+= max_linkhdr
;
1160 mhip
= mtod(m
, struct ip
*);
1162 if (hlen
> sizeof(struct ip
)) {
1163 mhlen
= ip_optcopy(ip
, mhip
) + sizeof(struct ip
);
1164 mhip
->ip_vhl
= IP_MAKE_VHL(IPVERSION
, mhlen
>> 2);
1167 /* XXX do we need to add ip->ip_off below ? */
1168 mhip
->ip_off
= ((off
- hlen
) >> 3) + ip
->ip_off
;
1169 if (off
+ len
>= ip
->ip_len
) { /* last fragment */
1170 len
= ip
->ip_len
- off
;
1171 m
->m_flags
|= M_LASTFRAG
;
1173 mhip
->ip_off
|= IP_MF
;
1174 mhip
->ip_len
= htons((u_short
)(len
+ mhlen
));
1175 m
->m_next
= m_copy(m0
, off
, len
);
1176 if (m
->m_next
== NULL
) { /* copy failed */
1178 error
= ENOBUFS
; /* ??? */
1179 ipstat
.ips_odropped
++;
1182 m
->m_pkthdr
.len
= mhlen
+ len
;
1183 m
->m_pkthdr
.rcvif
= NULL
;
1184 m
->m_pkthdr
.csum_flags
= m0
->m_pkthdr
.csum_flags
;
1185 mhip
->ip_off
= htons(mhip
->ip_off
);
1187 if (sw_csum
& CSUM_DELAY_IP
)
1188 mhip
->ip_sum
= in_cksum(m
, mhlen
);
1190 mnext
= &m
->m_nextpkt
;
1192 ipstat
.ips_ofragments
+= nfrags
;
1194 /* set first marker for fragment chain */
1195 m0
->m_flags
|= M_FIRSTFRAG
| M_FRAG
;
1196 m0
->m_pkthdr
.csum_data
= nfrags
;
1199 * Update first fragment by trimming what's been copied out
1200 * and updating header.
1202 m_adj(m0
, hlen
+ firstlen
- ip
->ip_len
);
1203 m0
->m_pkthdr
.len
= hlen
+ firstlen
;
1204 ip
->ip_len
= htons((u_short
)m0
->m_pkthdr
.len
);
1205 ip
->ip_off
|= IP_MF
;
1206 ip
->ip_off
= htons(ip
->ip_off
);
1208 if (sw_csum
& CSUM_DELAY_IP
)
1209 ip
->ip_sum
= in_cksum(m0
, hlen
);
1217 in_delayed_cksum(struct mbuf
*m
)
1220 u_short csum
, offset
;
1222 ip
= mtod(m
, struct ip
*);
1223 offset
= IP_VHL_HL(ip
->ip_vhl
) << 2 ;
1224 csum
= in_cksum_skip(m
, ip
->ip_len
, offset
);
1225 if (m
->m_pkthdr
.csum_flags
& CSUM_UDP
&& csum
== 0)
1227 offset
+= m
->m_pkthdr
.csum_data
; /* checksum offset */
1229 if (offset
+ sizeof(u_short
) > m
->m_len
) {
1230 kprintf("delayed m_pullup, m->len: %d off: %d p: %d\n",
1231 m
->m_len
, offset
, ip
->ip_p
);
1234 * this shouldn't happen, but if it does, the
1235 * correct behavior may be to insert the checksum
1236 * in the existing chain instead of rearranging it.
1238 m
= m_pullup(m
, offset
+ sizeof(u_short
));
1240 *(u_short
*)(m
->m_data
+ offset
) = csum
;
1244 * Insert IP options into preformed packet.
1245 * Adjust IP destination as required for IP source routing,
1246 * as indicated by a non-zero in_addr at the start of the options.
1248 * XXX This routine assumes that the packet has no options in place.
1250 static struct mbuf
*
1251 ip_insertoptions(struct mbuf
*m
, struct mbuf
*opt
, int *phlen
)
1253 struct ipoption
*p
= mtod(opt
, struct ipoption
*);
1255 struct ip
*ip
= mtod(m
, struct ip
*);
1258 optlen
= opt
->m_len
- sizeof p
->ipopt_dst
;
1259 if (optlen
+ (u_short
)ip
->ip_len
> IP_MAXPACKET
) {
1261 return (m
); /* XXX should fail */
1263 if (p
->ipopt_dst
.s_addr
)
1264 ip
->ip_dst
= p
->ipopt_dst
;
1265 if (m
->m_flags
& M_EXT
|| m
->m_data
- optlen
< m
->m_pktdat
) {
1266 MGETHDR(n
, MB_DONTWAIT
, MT_HEADER
);
1271 n
->m_pkthdr
.rcvif
= NULL
;
1272 n
->m_pkthdr
.len
= m
->m_pkthdr
.len
+ optlen
;
1273 m
->m_len
-= sizeof(struct ip
);
1274 m
->m_data
+= sizeof(struct ip
);
1277 m
->m_len
= optlen
+ sizeof(struct ip
);
1278 m
->m_data
+= max_linkhdr
;
1279 memcpy(mtod(m
, void *), ip
, sizeof(struct ip
));
1281 m
->m_data
-= optlen
;
1283 m
->m_pkthdr
.len
+= optlen
;
1284 ovbcopy(ip
, mtod(m
, caddr_t
), sizeof(struct ip
));
1286 ip
= mtod(m
, struct ip
*);
1287 bcopy(p
->ipopt_list
, ip
+ 1, optlen
);
1288 *phlen
= sizeof(struct ip
) + optlen
;
1289 ip
->ip_vhl
= IP_MAKE_VHL(IPVERSION
, *phlen
>> 2);
1290 ip
->ip_len
+= optlen
;
1295 * Copy options from ip to jp,
1296 * omitting those not copied during fragmentation.
1299 ip_optcopy(struct ip
*ip
, struct ip
*jp
)
1302 int opt
, optlen
, cnt
;
1304 cp
= (u_char
*)(ip
+ 1);
1305 dp
= (u_char
*)(jp
+ 1);
1306 cnt
= (IP_VHL_HL(ip
->ip_vhl
) << 2) - sizeof(struct ip
);
1307 for (; cnt
> 0; cnt
-= optlen
, cp
+= optlen
) {
1309 if (opt
== IPOPT_EOL
)
1311 if (opt
== IPOPT_NOP
) {
1312 /* Preserve for IP mcast tunnel's LSRR alignment. */
1318 KASSERT(cnt
>= IPOPT_OLEN
+ sizeof *cp
,
1319 ("ip_optcopy: malformed ipv4 option"));
1320 optlen
= cp
[IPOPT_OLEN
];
1321 KASSERT(optlen
>= IPOPT_OLEN
+ sizeof *cp
&& optlen
<= cnt
,
1322 ("ip_optcopy: malformed ipv4 option"));
1324 /* bogus lengths should have been caught by ip_dooptions */
1327 if (IPOPT_COPIED(opt
)) {
1328 bcopy(cp
, dp
, optlen
);
1332 for (optlen
= dp
- (u_char
*)(jp
+1); optlen
& 0x3; optlen
++)
1338 * IP socket option processing.
1341 ip_ctloutput(netmsg_t msg
)
1343 struct socket
*so
= msg
->base
.nm_so
;
1344 struct sockopt
*sopt
= msg
->ctloutput
.nm_sopt
;
1345 struct inpcb
*inp
= so
->so_pcb
;
1349 if (sopt
->sopt_level
!= IPPROTO_IP
) {
1354 switch (sopt
->sopt_dir
) {
1356 switch (sopt
->sopt_name
) {
1363 if (sopt
->sopt_valsize
> MLEN
) {
1367 MGET(m
, sopt
->sopt_td
? MB_WAIT
: MB_DONTWAIT
, MT_HEADER
);
1372 m
->m_len
= sopt
->sopt_valsize
;
1373 error
= soopt_to_kbuf(sopt
, mtod(m
, void *), m
->m_len
,
1375 error
= ip_pcbopts(sopt
->sopt_name
,
1376 &inp
->inp_options
, m
);
1384 case IP_RECVRETOPTS
:
1385 case IP_RECVDSTADDR
:
1389 error
= soopt_to_kbuf(sopt
, &optval
, sizeof optval
,
1393 switch (sopt
->sopt_name
) {
1395 inp
->inp_ip_tos
= optval
;
1399 inp
->inp_ip_ttl
= optval
;
1402 if (optval
>= 0 && optval
<= MAXTTL
)
1403 inp
->inp_ip_minttl
= optval
;
1407 #define OPTSET(bit) \
1409 inp->inp_flags |= bit; \
1411 inp->inp_flags &= ~bit;
1414 OPTSET(INP_RECVOPTS
);
1417 case IP_RECVRETOPTS
:
1418 OPTSET(INP_RECVRETOPTS
);
1421 case IP_RECVDSTADDR
:
1422 OPTSET(INP_RECVDSTADDR
);
1430 OPTSET(INP_RECVTTL
);
1440 case IP_MULTICAST_IF
:
1441 case IP_MULTICAST_VIF
:
1442 case IP_MULTICAST_TTL
:
1443 case IP_MULTICAST_LOOP
:
1444 case IP_ADD_MEMBERSHIP
:
1445 case IP_DROP_MEMBERSHIP
:
1446 error
= ip_setmoptions(sopt
, &inp
->inp_moptions
);
1450 error
= soopt_to_kbuf(sopt
, &optval
, sizeof optval
,
1456 case IP_PORTRANGE_DEFAULT
:
1457 inp
->inp_flags
&= ~(INP_LOWPORT
);
1458 inp
->inp_flags
&= ~(INP_HIGHPORT
);
1461 case IP_PORTRANGE_HIGH
:
1462 inp
->inp_flags
&= ~(INP_LOWPORT
);
1463 inp
->inp_flags
|= INP_HIGHPORT
;
1466 case IP_PORTRANGE_LOW
:
1467 inp
->inp_flags
&= ~(INP_HIGHPORT
);
1468 inp
->inp_flags
|= INP_LOWPORT
;
1477 #if defined(IPSEC) || defined(FAST_IPSEC)
1478 case IP_IPSEC_POLICY
:
1486 if ((error
= soopt_getm(sopt
, &m
)) != 0) /* XXX */
1488 soopt_to_mbuf(sopt
, m
);
1489 priv
= (sopt
->sopt_td
!= NULL
&&
1490 priv_check(sopt
->sopt_td
, PRIV_ROOT
) != 0) ? 0 : 1;
1491 req
= mtod(m
, caddr_t
);
1493 optname
= sopt
->sopt_name
;
1494 error
= ipsec4_set_policy(inp
, optname
, req
, len
, priv
);
1501 error
= ENOPROTOOPT
;
1507 switch (sopt
->sopt_name
) {
1510 if (inp
->inp_options
)
1511 soopt_from_kbuf(sopt
, mtod(inp
->inp_options
,
1513 inp
->inp_options
->m_len
);
1515 sopt
->sopt_valsize
= 0;
1522 case IP_RECVRETOPTS
:
1523 case IP_RECVDSTADDR
:
1528 switch (sopt
->sopt_name
) {
1531 optval
= inp
->inp_ip_tos
;
1535 optval
= inp
->inp_ip_ttl
;
1538 optval
= inp
->inp_ip_minttl
;
1541 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1544 optval
= OPTBIT(INP_RECVOPTS
);
1547 case IP_RECVRETOPTS
:
1548 optval
= OPTBIT(INP_RECVRETOPTS
);
1551 case IP_RECVDSTADDR
:
1552 optval
= OPTBIT(INP_RECVDSTADDR
);
1556 optval
= OPTBIT(INP_RECVTTL
);
1560 optval
= OPTBIT(INP_RECVIF
);
1564 if (inp
->inp_flags
& INP_HIGHPORT
)
1565 optval
= IP_PORTRANGE_HIGH
;
1566 else if (inp
->inp_flags
& INP_LOWPORT
)
1567 optval
= IP_PORTRANGE_LOW
;
1573 optval
= OPTBIT(INP_FAITH
);
1576 soopt_from_kbuf(sopt
, &optval
, sizeof optval
);
1579 case IP_MULTICAST_IF
:
1580 case IP_MULTICAST_VIF
:
1581 case IP_MULTICAST_TTL
:
1582 case IP_MULTICAST_LOOP
:
1583 case IP_ADD_MEMBERSHIP
:
1584 case IP_DROP_MEMBERSHIP
:
1585 error
= ip_getmoptions(sopt
, inp
->inp_moptions
);
1588 #if defined(IPSEC) || defined(FAST_IPSEC)
1589 case IP_IPSEC_POLICY
:
1591 struct mbuf
*m
= NULL
;
1596 req
= mtod(m
, caddr_t
);
1599 error
= ipsec4_get_policy(so
->so_pcb
, req
, len
, &m
);
1601 error
= soopt_from_mbuf(sopt
, m
); /* XXX */
1609 error
= ENOPROTOOPT
;
1615 lwkt_replymsg(&msg
->lmsg
, error
);
1619 * Set up IP options in pcb for insertion in output packets.
1620 * Store in mbuf with pointer in pcbopt, adding pseudo-option
1621 * with destination address if source routed.
1624 ip_pcbopts(int optname
, struct mbuf
**pcbopt
, struct mbuf
*m
)
1630 /* turn off any old options */
1634 if (m
== NULL
|| m
->m_len
== 0) {
1636 * Only turning off any previous options.
1643 if (m
->m_len
% sizeof(int32_t))
1646 * IP first-hop destination address will be stored before
1647 * actual options; move other options back
1648 * and clear it when none present.
1650 if (m
->m_data
+ m
->m_len
+ sizeof(struct in_addr
) >= &m
->m_dat
[MLEN
])
1653 m
->m_len
+= sizeof(struct in_addr
);
1654 cp
= mtod(m
, u_char
*) + sizeof(struct in_addr
);
1655 ovbcopy(mtod(m
, caddr_t
), cp
, cnt
);
1656 bzero(mtod(m
, caddr_t
), sizeof(struct in_addr
));
1658 for (; cnt
> 0; cnt
-= optlen
, cp
+= optlen
) {
1659 opt
= cp
[IPOPT_OPTVAL
];
1660 if (opt
== IPOPT_EOL
)
1662 if (opt
== IPOPT_NOP
)
1665 if (cnt
< IPOPT_OLEN
+ sizeof *cp
)
1667 optlen
= cp
[IPOPT_OLEN
];
1668 if (optlen
< IPOPT_OLEN
+ sizeof *cp
|| optlen
> cnt
)
1679 * user process specifies route as:
1681 * D must be our final destination (but we can't
1682 * check that since we may not have connected yet).
1683 * A is first hop destination, which doesn't appear in
1684 * actual IP option, but is stored before the options.
1686 if (optlen
< IPOPT_MINOFF
- 1 + sizeof(struct in_addr
))
1688 m
->m_len
-= sizeof(struct in_addr
);
1689 cnt
-= sizeof(struct in_addr
);
1690 optlen
-= sizeof(struct in_addr
);
1691 cp
[IPOPT_OLEN
] = optlen
;
1693 * Move first hop before start of options.
1695 bcopy(&cp
[IPOPT_OFFSET
+1], mtod(m
, caddr_t
),
1696 sizeof(struct in_addr
));
1698 * Then copy rest of options back
1699 * to close up the deleted entry.
1701 ovbcopy(&cp
[IPOPT_OFFSET
+1] + sizeof(struct in_addr
),
1702 &cp
[IPOPT_OFFSET
+1],
1703 cnt
- (IPOPT_MINOFF
- 1));
1707 if (m
->m_len
> MAX_IPOPTLEN
+ sizeof(struct in_addr
))
1719 * The whole multicast option thing needs to be re-thought.
1720 * Several of these options are equally applicable to non-multicast
1721 * transmission, and one (IP_MULTICAST_TTL) totally duplicates a
1722 * standard option (IP_TTL).
1726 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index.
1728 static struct ifnet
*
1729 ip_multicast_if(struct in_addr
*a
, int *ifindexp
)
1736 if (ntohl(a
->s_addr
) >> 24 == 0) {
1737 ifindex
= ntohl(a
->s_addr
) & 0xffffff;
1738 if (ifindex
< 0 || if_index
< ifindex
)
1740 ifp
= ifindex2ifnet
[ifindex
];
1742 *ifindexp
= ifindex
;
1744 ifp
= INADDR_TO_IFP(a
);
1750 * Set the IP multicast options in response to user setsockopt().
1753 ip_setmoptions(struct sockopt
*sopt
, struct ip_moptions
**imop
)
1757 struct in_addr addr
;
1758 struct ip_mreq mreq
;
1760 struct ip_moptions
*imo
= *imop
;
1765 * No multicast option buffer attached to the pcb;
1766 * allocate one and initialize to default values.
1768 imo
= kmalloc(sizeof *imo
, M_IPMOPTS
, M_WAITOK
);
1771 imo
->imo_multicast_ifp
= NULL
;
1772 imo
->imo_multicast_addr
.s_addr
= INADDR_ANY
;
1773 imo
->imo_multicast_vif
= -1;
1774 imo
->imo_multicast_ttl
= IP_DEFAULT_MULTICAST_TTL
;
1775 imo
->imo_multicast_loop
= IP_DEFAULT_MULTICAST_LOOP
;
1776 imo
->imo_num_memberships
= 0;
1778 switch (sopt
->sopt_name
) {
1779 /* store an index number for the vif you wanna use in the send */
1780 case IP_MULTICAST_VIF
:
1781 if (legal_vif_num
== 0) {
1785 error
= soopt_to_kbuf(sopt
, &i
, sizeof i
, sizeof i
);
1788 if (!legal_vif_num(i
) && (i
!= -1)) {
1792 imo
->imo_multicast_vif
= i
;
1795 case IP_MULTICAST_IF
:
1797 * Select the interface for outgoing multicast packets.
1799 error
= soopt_to_kbuf(sopt
, &addr
, sizeof addr
, sizeof addr
);
1804 * INADDR_ANY is used to remove a previous selection.
1805 * When no interface is selected, a default one is
1806 * chosen every time a multicast packet is sent.
1808 if (addr
.s_addr
== INADDR_ANY
) {
1809 imo
->imo_multicast_ifp
= NULL
;
1813 * The selected interface is identified by its local
1814 * IP address. Find the interface and confirm that
1815 * it supports multicasting.
1818 ifp
= ip_multicast_if(&addr
, &ifindex
);
1819 if (ifp
== NULL
|| !(ifp
->if_flags
& IFF_MULTICAST
)) {
1821 error
= EADDRNOTAVAIL
;
1824 imo
->imo_multicast_ifp
= ifp
;
1826 imo
->imo_multicast_addr
= addr
;
1828 imo
->imo_multicast_addr
.s_addr
= INADDR_ANY
;
1832 case IP_MULTICAST_TTL
:
1834 * Set the IP time-to-live for outgoing multicast packets.
1835 * The original multicast API required a char argument,
1836 * which is inconsistent with the rest of the socket API.
1837 * We allow either a char or an int.
1839 if (sopt
->sopt_valsize
== 1) {
1841 error
= soopt_to_kbuf(sopt
, &ttl
, 1, 1);
1844 imo
->imo_multicast_ttl
= ttl
;
1847 error
= soopt_to_kbuf(sopt
, &ttl
, sizeof ttl
, sizeof ttl
);
1853 imo
->imo_multicast_ttl
= ttl
;
1857 case IP_MULTICAST_LOOP
:
1859 * Set the loopback flag for outgoing multicast packets.
1860 * Must be zero or one. The original multicast API required a
1861 * char argument, which is inconsistent with the rest
1862 * of the socket API. We allow either a char or an int.
1864 if (sopt
->sopt_valsize
== 1) {
1867 error
= soopt_to_kbuf(sopt
, &loop
, 1, 1);
1870 imo
->imo_multicast_loop
= !!loop
;
1874 error
= soopt_to_kbuf(sopt
, &loop
, sizeof loop
,
1878 imo
->imo_multicast_loop
= !!loop
;
1882 case IP_ADD_MEMBERSHIP
:
1884 * Add a multicast group membership.
1885 * Group must be a valid IP multicast address.
1887 error
= soopt_to_kbuf(sopt
, &mreq
, sizeof mreq
, sizeof mreq
);
1891 if (!IN_MULTICAST(ntohl(mreq
.imr_multiaddr
.s_addr
))) {
1897 * If no interface address was provided, use the interface of
1898 * the route to the given multicast address.
1900 if (mreq
.imr_interface
.s_addr
== INADDR_ANY
) {
1901 struct sockaddr_in dst
;
1904 bzero(&dst
, sizeof(struct sockaddr_in
));
1905 dst
.sin_len
= sizeof(struct sockaddr_in
);
1906 dst
.sin_family
= AF_INET
;
1907 dst
.sin_addr
= mreq
.imr_multiaddr
;
1908 rt
= rtlookup((struct sockaddr
*)&dst
);
1910 error
= EADDRNOTAVAIL
;
1917 ifp
= ip_multicast_if(&mreq
.imr_interface
, NULL
);
1921 * See if we found an interface, and confirm that it
1922 * supports multicast.
1924 if (ifp
== NULL
|| !(ifp
->if_flags
& IFF_MULTICAST
)) {
1925 error
= EADDRNOTAVAIL
;
1930 * See if the membership already exists or if all the
1931 * membership slots are full.
1933 for (i
= 0; i
< imo
->imo_num_memberships
; ++i
) {
1934 if (imo
->imo_membership
[i
]->inm_ifp
== ifp
&&
1935 imo
->imo_membership
[i
]->inm_addr
.s_addr
1936 == mreq
.imr_multiaddr
.s_addr
)
1939 if (i
< imo
->imo_num_memberships
) {
1944 if (i
== IP_MAX_MEMBERSHIPS
) {
1945 error
= ETOOMANYREFS
;
1950 * Everything looks good; add a new record to the multicast
1951 * address list for the given interface.
1953 if ((imo
->imo_membership
[i
] =
1954 in_addmulti(&mreq
.imr_multiaddr
, ifp
)) == NULL
) {
1959 ++imo
->imo_num_memberships
;
1963 case IP_DROP_MEMBERSHIP
:
1965 * Drop a multicast group membership.
1966 * Group must be a valid IP multicast address.
1968 error
= soopt_to_kbuf(sopt
, &mreq
, sizeof mreq
, sizeof mreq
);
1972 if (!IN_MULTICAST(ntohl(mreq
.imr_multiaddr
.s_addr
))) {
1979 * If an interface address was specified, get a pointer
1980 * to its ifnet structure.
1982 if (mreq
.imr_interface
.s_addr
== INADDR_ANY
)
1985 ifp
= ip_multicast_if(&mreq
.imr_interface
, NULL
);
1987 error
= EADDRNOTAVAIL
;
1993 * Find the membership in the membership array.
1995 for (i
= 0; i
< imo
->imo_num_memberships
; ++i
) {
1997 imo
->imo_membership
[i
]->inm_ifp
== ifp
) &&
1998 imo
->imo_membership
[i
]->inm_addr
.s_addr
==
1999 mreq
.imr_multiaddr
.s_addr
)
2002 if (i
== imo
->imo_num_memberships
) {
2003 error
= EADDRNOTAVAIL
;
2008 * Give up the multicast address record to which the
2009 * membership points.
2011 in_delmulti(imo
->imo_membership
[i
]);
2013 * Remove the gap in the membership array.
2015 for (++i
; i
< imo
->imo_num_memberships
; ++i
)
2016 imo
->imo_membership
[i
-1] = imo
->imo_membership
[i
];
2017 --imo
->imo_num_memberships
;
2027 * If all options have default values, no need to keep the mbuf.
2029 if (imo
->imo_multicast_ifp
== NULL
&&
2030 imo
->imo_multicast_vif
== -1 &&
2031 imo
->imo_multicast_ttl
== IP_DEFAULT_MULTICAST_TTL
&&
2032 imo
->imo_multicast_loop
== IP_DEFAULT_MULTICAST_LOOP
&&
2033 imo
->imo_num_memberships
== 0) {
2034 kfree(*imop
, M_IPMOPTS
);
2042 * Return the IP multicast options in response to user getsockopt().
2045 ip_getmoptions(struct sockopt
*sopt
, struct ip_moptions
*imo
)
2047 struct in_addr addr
;
2048 struct in_ifaddr
*ia
;
2053 switch (sopt
->sopt_name
) {
2054 case IP_MULTICAST_VIF
:
2056 optval
= imo
->imo_multicast_vif
;
2059 soopt_from_kbuf(sopt
, &optval
, sizeof optval
);
2062 case IP_MULTICAST_IF
:
2063 if (imo
== NULL
|| imo
->imo_multicast_ifp
== NULL
)
2064 addr
.s_addr
= INADDR_ANY
;
2065 else if (imo
->imo_multicast_addr
.s_addr
) {
2066 /* return the value user has set */
2067 addr
= imo
->imo_multicast_addr
;
2069 ia
= IFP_TO_IA(imo
->imo_multicast_ifp
);
2070 addr
.s_addr
= (ia
== NULL
) ? INADDR_ANY
2071 : IA_SIN(ia
)->sin_addr
.s_addr
;
2073 soopt_from_kbuf(sopt
, &addr
, sizeof addr
);
2076 case IP_MULTICAST_TTL
:
2078 optval
= coptval
= IP_DEFAULT_MULTICAST_TTL
;
2080 optval
= coptval
= imo
->imo_multicast_ttl
;
2081 if (sopt
->sopt_valsize
== 1)
2082 soopt_from_kbuf(sopt
, &coptval
, 1);
2084 soopt_from_kbuf(sopt
, &optval
, sizeof optval
);
2087 case IP_MULTICAST_LOOP
:
2089 optval
= coptval
= IP_DEFAULT_MULTICAST_LOOP
;
2091 optval
= coptval
= imo
->imo_multicast_loop
;
2092 if (sopt
->sopt_valsize
== 1)
2093 soopt_from_kbuf(sopt
, &coptval
, 1);
2095 soopt_from_kbuf(sopt
, &optval
, sizeof optval
);
2099 error
= ENOPROTOOPT
;
2106 * Discard the IP multicast options.
2109 ip_freemoptions(struct ip_moptions
*imo
)
2114 for (i
= 0; i
< imo
->imo_num_memberships
; ++i
)
2115 in_delmulti(imo
->imo_membership
[i
]);
2116 kfree(imo
, M_IPMOPTS
);
2121 * Routine called from ip_output() to loop back a copy of an IP multicast
2122 * packet to the input queue of a specified interface. Note that this
2123 * calls the output routine of the loopback "driver", but with an interface
2124 * pointer that might NOT be a loopback interface -- evil, but easier than
2125 * replicating that code here.
2128 ip_mloopback(struct ifnet
*ifp
, struct mbuf
*m
, struct sockaddr_in
*dst
,
2134 copym
= m_copypacket(m
, MB_DONTWAIT
);
2135 if (copym
!= NULL
&& (copym
->m_flags
& M_EXT
|| copym
->m_len
< hlen
))
2136 copym
= m_pullup(copym
, hlen
);
2137 if (copym
!= NULL
) {
2139 * if the checksum hasn't been computed, mark it as valid
2141 if (copym
->m_pkthdr
.csum_flags
& CSUM_DELAY_DATA
) {
2142 in_delayed_cksum(copym
);
2143 copym
->m_pkthdr
.csum_flags
&= ~CSUM_DELAY_DATA
;
2144 copym
->m_pkthdr
.csum_flags
|=
2145 CSUM_DATA_VALID
| CSUM_PSEUDO_HDR
;
2146 copym
->m_pkthdr
.csum_data
= 0xffff;
2149 * We don't bother to fragment if the IP length is greater
2150 * than the interface's MTU. Can this possibly matter?
2152 ip
= mtod(copym
, struct ip
*);
2153 ip
->ip_len
= htons(ip
->ip_len
);
2154 ip
->ip_off
= htons(ip
->ip_off
);
2156 if (ip
->ip_vhl
== IP_VHL_BORING
) {
2157 ip
->ip_sum
= in_cksum_hdr(ip
);
2159 ip
->ip_sum
= in_cksum(copym
, hlen
);
2163 * It's not clear whether there are any lingering
2164 * reentrancy problems in other areas which might
2165 * be exposed by using ip_input directly (in
2166 * particular, everything which modifies the packet
2167 * in-place). Yet another option is using the
2168 * protosw directly to deliver the looped back
2169 * packet. For the moment, we'll err on the side
2170 * of safety by using if_simloop().
2173 if (dst
->sin_family
!= AF_INET
) {
2174 kprintf("ip_mloopback: bad address family %d\n",
2176 dst
->sin_family
= AF_INET
;
2179 get_mplock(); /* is if_simloop() mpsafe yet? */
2180 if_simloop(ifp
, copym
, dst
->sin_family
, 0);